A comparison of interpolation techniques for RR interval fitting in AR spectrum estimation

In this paper, we have compared basic interpolation techniques (linear interpolation, Lagrange interpolation, Hermite interpolation, and cubic spline interpolation) to find the optimum method for RR interval fitting in heart rate variability (HRV) analysis. It is required that a sequence of RR intervals have to be resampled to make it as if it is a regularly sampled signal since the input signal has to be satisfied a steady state assumption for frequency domain analysis. Several interpolation techniques have been applied to cope with this problem. To find the optimum algorithm among them, we have compared the algorithms in terms of processing times and error rates of HRV parameters (normalized low frequency (LF_norm), normalized high frequency (HF_norm), LF/HF ratio). We have also employed EUROBAVAR datasets which include 10-12 min recorded RR interval data for the experiment. From the experiment, we can notice that the Lagrange interpolation technique with order of 3 is the most appropriate algorithm for the RR interval fitting in the autoregressive spectrum estimation since it requires low processing time (0.028 seconds in the Intel Core 2 Quad @ 2.40 GHz desktop computer) and shows the lowest error rates in HRV parameter calculation.